Tissue absence initiates regeneration through Follistatin-mediated inhibition of Activin signaling

1Regeneration occurs in widespread contexts and species. Invertebrates such as Hydra are capable of whole-animal regeneration from tissue fragments, and many vertebrates can regenerate appendages or repair damaged organs. Despite this widespread relevance, the central mechanisms that drive regeneration are poorly understood.

2Planarians are flatworms capable of regeneration following an almost limitless variety of injuries and have emerged as a powerful model for exploring the molecular underpinnings of regeneration. New tissues are formed at planarian wound sites in a process called blastema formation, and pre-existing tissues are reorganized after amputation to accommodate reduced animal size and further generate missing tissues. The source of regenerated tissue in planarians is a population of adult dividing cells called neoblasts, which include pluripotent stem cells called clonogenic neoblasts (cNeoblasts). Neoblasts are the only somatic cycling cells in adult animals and can be specifically ablated by gamma irradiation, allowing for dissection of the requirements for neoblasts in regenerative processes. Recent work has described the earliest molecular and cellular events that occur following injury. One finding to emerge from this work is that animals initiate distinct cellular and molecular responses to ‘major injuries’ that remove significant amounts of tissue (e.g., head amputation) and to ‘simple injuries’ that require only minimal healing for repair (wounds that do not elicit blastema formation, such as punctures or incisions). Following simple injury, for example, animals display an increase in mitotic numbers 6 hr after injury before returning to baseline levels, and expression of numerous wound-induced genes becomes undetectable by 24 hr after injury. Following a major injury, these same initial responses are observed, but subsequent responses are also activated: the 6 hr increase in mitotic numbers is followed by a second increase 48 hr after amputation, and wound-induced gene expression persists beyond 24 hr and is refined over several days. These responses are referred to as the ‘missing-tissue response’. How animals distinguish between injuries involving varying amounts of tissue loss and regulate these distinct wound response programs remains unknown.

3We identified Smed-follistatin as required for molecular and cellular ‘missing-tissue’ responses during regeneration. Specifically, Follistatin-mediated inhibition of Activin signaling is required for regeneration to occur, with Smed-follistatin expression at wounds controlled by the extent of tissue absence following injury. These results suggest a mechanism by which regenerative responses can be initiated.

4To identify genes mediating regeneration-specific wound responses, we inhibited recently identified wound-induced genes with RNA interference (RNAi). Inhibition of Smed-follistatin (follistatin or fst), a gene encoding a Follistatin-like TGF-β-superfamily inhibitor, completely blocked regeneration (Figure 1A, Figure 1—figure supplement 1). No brain regeneration or anterior pole regeneration was observed in fst(RNAi) animals (Figure 1A, Figure 1—figure supplement 2). The anterior pole phenotype is consistent with a described role for follistatin in anterior regeneration. fst(RNAi) animals, however, also failed to produce a blastema following either tail amputation or the excision of lateral tissue wedges that left anterior and posterior poles intact (Figure 1B). These data demonstrate that fst is required broadly for regeneration.

5Planarians constantly maintain adult tissues through cell turnover involving neoblasts. Consequently, most genes required for regeneration are also required for tissue turnover because of an involvement in neoblast biology. Strikingly, unamputated fst(RNAi) animals did not shrink or lose structures, as is typically seen in animals with neoblast dysfunction, even after several months of significant expression reduction with RNAi (Figure 1C, Figure 1—figure supplement 3). Furthermore, amputated animals—despite failing to regenerate—displayed ongoing long-term neoblast-based tissue turnover of remaining tissue (Figure 1D). Together, these data suggest that the requirement for fst in tissue replacement is specific to regeneration, as it is not detectably required for neoblast-mediated tissue turnover. Because of the rarity of genes required for regeneration but not tissue turnover, fst was a good candidate for specifically mediating the processes that occur following injury to bring about regeneration.

6fst expression was induced at wounds by 6 hr following amputation and persisted for several days, with maximal expression around 12 hr post-amputation (Figure 1E, Figure 1—figure supplement 4). In unamputated animals, fst was expressed sparsely throughout the animal, including ventrally, in a thin peripheral domain, and at the anterior pole (Figure 1E, Figure 1—figure supplement 4; ). Injection of fst dsRNA only after amputation caused poor blastema formation and regeneration defects (Figure 1—figure supplement 5), consistent with a requirement for wound-induced fst expression in regeneration. We conclude that fst is a wound-induced factor required for regeneration.

7To characterize the defects underlying regeneration failure in fst(RNAi) animals, we first investigated whether fst regulates neoblast function in regeneration. Neoblasts can be visualized by detecting neoblast-specific transcripts through whole-mount in situ hybridization and quantified using flow cytometry. fst(RNAi) animals displayed normal neoblast numbers prior to amputation, indicating that the observed regeneration failure is not caused by neoblast loss (Figure 2A). We next assessed whether neoblasts respond to injury in fst(RNAi) animals. The neoblast response to injury involves two peaks (6 hr and 48 hr post-amputation) in mitotic cell numbers, in between which neoblasts migrate to wounds. The first peak is generically induced by all injury types and is spatially widespread. The second peak occurs specifically following major injuries (removing tissues) and is biased toward wound sites. Amputated fst(RNAi) animals displayed a normal 6 hr mitotic peak, indicating that a normal generic injury response was present (Figure 2B). By contrast, these animals failed to display a 48 hr mitotic peak (Figure 2B). fst(RNAi) animals did however display localization of mitoses toward wound sites 48 hr after amputation (Figure 2—figure supplement 1), and neoblast enrichment at wound sites 18 hr after injury (Figure 2C), indicating that neoblast migration occurred normally.

8Given that fst(RNAi) animals displayed a defective proliferative response to missing tissue, we tested whether these animals produced regenerative progenitor cell types. Head amputation normally induces neoblasts to produce ovo+ eye progenitors, but this process failed in fst(RNAi) animals (Figure 2D). From these data taken together, we conclude that fst is required for several aspects of the regeneration-specific neoblast response to injury.

9The abnormal missing-tissue-specific mitotic response of fst(RNAi) animals raised the possibility that other missing tissue responses could also require fst. Apoptosis increases following injury in planarians, and, like the mitotic response, this increase involves a generic injury phase and a missing-tissue-specific phase. First, a local apoptosis burst occurs at wound sites 4 hr following any injury; second, a body-wide apoptosis burst occurs 72 hr after injury, but only in cases involving missing tissue. The apoptosis level in this latter phase scales with the amount of missing tissue. Planarians possess a centrally located pharynx used for feeding and defecation; measuring apoptotic cell numbers by TUNEL within the pharynx is an established assay for quantifying the body-wide increase in apoptosis that occurs 72 hr post-amputation. Strikingly, fst(RNAi) pharynges displayed little increase in apoptotic cell numbers 72 hr post amputation, whereas a roughly 20-fold increase from pre-amputation levels occurred in control pharynges (Figure 3A). fst(RNAi) animals had a normal 4 hr apoptosis burst, indicating that fst is not generally required for apoptosis (Figure 3B). The 72 hr apoptotic response occurs in animals that have had their neoblasts ablated and cannot regenerate. Therefore, the failure of fst(RNAi) animals to produce this response cannot be explained as a non-specific result of regeneration failure.

10In addition to the cellular responses to missing tissue described above, persistence of wound-induced gene expression is another aspect of the planarian missing-tissue response. We observed less expression of two wound-response genes in fst(RNAi) animals than in controls 24–48 hr post-amputation, despite expression levels being indistinguishable at earlier timepoints (Figure 3C). Notably, some wound-induced genes display expression that inversely scales with missing tissue amount; for example, Smed-delta-1 displays higher expression after an incision or puncture (simple injuries) than after amputation (a major injury). Amputated fst(RNAi) animals displayed a higher, rather than lower, level of Smed-delta-1 expression than did controls 24 hr after amputation (Figure 3D). Therefore, the lower expression levels observed for other wound-induced genes in fst(RNAi) animals do not reflect generically lower gene expression at wounds, but instead a specific requirement for fst for missing-tissue-specific gene expression.

11Irradiated animals (which cannot regenerate) can display either higher or lower levels of wound-induced expression, depending on the gene examined. Indeed, some wound-induced genes were similarly affected between irradiated and fst(RNAi) animals, while others were oppositely affected (Figure 3—figure supplement 1). As was the case for the failed apoptotic response of fst(RNAi) animals, the missing-tissue gene expression defects of fst(RNAi) animals cannot therefore be explained as a side-effect of regenerative failure.

12In addition to producing a regeneration blastema, amputated animals must reorganize and rescale remaining tissue in a process termed morphallaxis. Some aspects of this process do not require blastema formation. For example, wntP-2 (also known as wnt11-5) is normally expressed in planarian tails and its expression domain restricts posteriorly within 48 hr of amputation whether regeneration proceeds or not. fst(RNAi) animals did not rescale the wntP-2 expression domain 48 hr following amputation, further supporting a model in which fst is required for responding to missing tissue (Figure 3E). Following head amputation, head fragments not only produce posterior-specific cell types but also reduce numbers of anterior-specific cell types (which are overabundant for the new fragment dimensions). This process failed in fst(RNAi) head fragments (Figure 3F). Finally, fst(RNAi) fragments did not produce pharynges de novo (which normally occurs in pre-existing head and tail fragment tissue) (Figure 3—figure supplement 1). By contrast, RNAi of a different gene that blocked blastema formation (smad1) did not block pharynx formation, indicating this defect is not a simple consequence of blastema formation failure (Figure 3—figure supplement 1). We conclude that fst is required broadly for missing-tissue-specific wound responses, and that these defects likely underlie the inability of fst(RNAi) animals to regenerate.

13Because Follistatin proteins are well-characterized extracellular inhibitors of TGF-β ligands, we sought to identify putative TGF-β ligands that Smed-Follistatin might regulate to promote regeneration. Seven putative TGF-β superfamily members exist in the Schmidtea mediterranea genome (Figure 4—figure supplement 1 and;;;;;; ). If Fst regulates one of the proteins encoded by these genes, then RNAi of that gene might suppress the fst RNAi phenotype. We tested this possibility (see Figure 4—figure supplement 2 and ‘Materials and methods’ for details) and found that RNAi of either of two genes, Smed-activin-1 (act-1 in short) or Smed-activin-2 (act-2), strongly suppressed the blastema formation defect (Figure 4A), the failure to regenerate a brain (Figure 4A), and the failed missing-tissue apoptotic response of fst(RNAi) animals (Figure 4B); RNAi of act-2 can also restore anterior pole regeneration in fst(RNAi) animals. Given that Follistatin proteins can directly regulate Activin proteins in other organisms, these data suggest that Follistatin promotes missing tissue responses by inhibiting the function of Activin proteins.

14Given that activin expression is required for the fst(RNAi) phenotype, we investigated the consequences of act-1 RNAi on regeneration. Although act-2(RNAi) has been reported to produce posterior regeneration defects, act-1(RNAi) animals were capable of regenerating (Figure 4—figure supplement 3, Figure 4—figure supplement 4) and, as with fst(RNAi), displayed normal neoblast turnover during homeostatic growth (Figure 4—figure supplement 5). act-1(RNAi) survived after amputation as well as controls did (observed more than a month following injury, n = 10/10). act-1(RNAi) animals did however display some abnormalities. Although act-1(RNAi) animals displayed normal ovo+ eye progenitor numbers prior to amputation, increased numbers as compared to controls were present following amputation (Figure 4C). By contrast, fst RNAi caused the opposite phenotype of reduced ovo+ eye progenitor formation. These data raise the possibility that act-1 regulates responses to injury, with some aspects of regeneration overactive following act-1 inhibition.

15Because fst is required for regeneration but not for normal tissue turnover, we reasoned that fst expression might be high following amputation, an injury type requiring significant tissue regeneration, but low following incision or puncture, injuries requiring only wound healing. We therefore assessed fst as compared to act expression at wounds following either incision or excision of a tissue wedge. Increased act-1 expression was not detected following either type of wound, with expression detected throughout the intestine of uninjured animals, suggesting an intestinal source of Activin-1 protein (Figure 5A). act-2 expression was similar to act-1 in intact animals, but unlike act-1 is wound-induced (Figure 5B, Figure 5—figure supplement 1; ). Indeed, act-2 was wound-induced following either incision or tissue wedge excision, with expression persisting for several days irrespective of injury severity (Figure 5C, Figure 5—figure supplement 2). By contrast, fst expression was induced at both wound types by 6 hr after injury, but by 48 hr after injury was present only at wedge excision wound sites (Figure 5C, Figure 5—figure supplement 2). These results indicate that fst expression persists longer at wounds that result in tissue absence. Furthermore, fst expression was greater at wounds involving a large amount of missing tissue (assessed at 48 hr) than at wounds with little missing tissue (Figure 5—figure supplement 3). Together, these data are consistent with a model in which wound-induced fst expression levels are regulated by the amount of missing tissue. In this model, fst promotes regenerative responses by inhibition of act-1 and act-2 following major injury (Figure 5D).

16All long-living animals face the prospect of injury and require regenerative mechanisms. Planarians are an exceptional example of the regenerative potential of animals. Distinct cellular and molecular programs for responding to simple injury vs missing tissue exist in planarians. In the case of injuries involving substantial missing tissue, animals mount unique mitotic and apoptotic responses and produce an extended program of wound-induced gene expression. These events represent the earliest described divergent behaviors following major injuries requiring regeneration vs simple injuries requiring only wound healing. A central question has therefore become how these distinct responses are mediated.

17We identified a gene encoding a homolog of the TGF-β inhibitor, follistatin, that is required for regeneration and for regeneration-specific cellular and molecular responses to injury. Our data suggest that inhibition of Activin signaling by Fst is required for initiating a regenerative response at wounds following major injury. Finally, fst is wound-induced, with the level of fst expression persisting at high levels longer following a major injury than following a simple injury. We propose that wound-induced fst expression allows for regenerative responses to be initiated specifically as a consequence of tissue absence.

18fst is the first gene known to be required for regeneration-specific responses in planarians. Not all missing-tissue responses are abolished following fst inhibition, however. For example, neoblast migration to amputation sites occurred normally in fst(RNAi) animals, despite the absence of a normal proliferative response. Similarly, although expression of act-1 and act-2 are required for the fst(RNAi) phenotype, inhibition of activin expression in the absence of amputation does not affect homeostatic tissue turnover or induce a regeneration-like state, demonstrating that the suppression of Activin alone is not sufficient to induce missing-tissue responses. Therefore, some aspects of the missing-tissue response to injury require an as yet unknown ‘missing-tissue’ signal or signals that operate independently of fst and Activin signaling. Identifying and characterizing these processes will be important for understanding how the decision to mount a regenerative response occurs.

19Our findings describe a system in which suppression of Activin signaling is required for regeneration. The possibility therefore exists that Activin signaling may serve similar functions in other organisms. Indeed, TGF-β signaling has been implicated as a negative regulator of regeneration in a variety of contexts, including following partial hepatectomy, in embryonic chick retinas, in renal regeneration following ischemia/reperfusion injuries, and for mouse skeletal muscle regeneration. Given the relevance of these systems to human medicine, it will be important to investigate to what extent regenerative regimes recapitulate the mechanisms observed in planarians. Interestingly, a number of systems use TGF-β signaling to promote rather than suppress regeneration: TGF-β signaling is involved in axolotl limb and Xenopus tail regeneration, activin expression can be induced by wounding and exogenous TGF-β can speed healing in mammals, TGF-β signaling can promote regeneration following mouse ear hole-punching, and wound-induced activin promotes cell proliferation and migration following zebrafish fin amputation. Despite these contextual differences, TGF-β signaling plays a major role in many forms of regeneration studied. Therefore, uncovering ‘missing-tissue’ signals in planarians, describing how these signals interact with Activin signaling, and identifying the key factors regulated by these signals will inform a broad understanding of core regenerative mechanisms.

20For RNA probes, genes were cloned into pGEM and amplified with T7-promoter-containing primers. For RNAi, genes were cloned into pPR244 as described. activin-1 was cloned with primers 5’-TCAACTGAAACGGAAGTTGG-3’ and 5’-TGGTGGATCCTTACTTGCAG-3’, activin-2 with primers 5’-ACCAATTATGGCCAATCCAG-3’ and 5’-CCGGCTAATTGTGAACAAAC-3’, and follistatin with 5’-CACAAGAGGCTGCAGTGAAT-3’ and 5’-CATTCAGAAGGCATTGTCCA-3’.

21The control dsRNA for all RNAi experiments was unc-22 from Caenorhabditis elegans. RNAi experiments were performed by feeding a mixture of liver and bacteria expressing dsRNA. 20 ml of bacterial culture was pelleted and resuspended in 60 μl of liver. For fst and act-1 RNAi regeneration experiments, animals were fed on day 0, day 4, day 8, and day 12, amputated on day 16/17 and either soaked for 6 hr in 1 µg/µl dsRNA (TUNEL experiments), soaked for 2 hr in dsRNA (gene expression experiments), or not soaked in dsRNA. For suppression experiments, totals from two separate experiments were pooled: (1) animals were fed fst dsRNA on day 0, day 4, day 8, and day 12, fed candidate gene dsRNA on day 16, day 20, and day 23, and amputated on day 24. (2) Animals were amputated and injected four times with a 30 nl equimolar mixture of fst and candidate gene dsRNA on day 0, injected without amputation on day 1, amputated and injected on day 4, and injected only on day 5. Animals were scored and fixed 8 days after the final amputation.

22Whole-mount in situ hybridizations and fluorescence in situ hybridizations (FISH) were performed as described. For double/triple labeling, HRP-inactivation was performed between labelings (4% formaldehyde, 30 min). Immunostainings were performed as previously described using tyramide signal enhancement. TUNEL was performed as previously described.

23For elimination of neoblasts, planarians were exposed to 6000 rad (6K, ∼72 min) using a cesium source (∼83 rad/min).

24Animals were amputated in cold CMFB, and cells prepared as described. For quantification of X1 cells, five animals were used per RNAi condition in triplicate. Analyses and sorting were performed using a Moflo3 FACS sorter (Dako-Cytomation, Carpinteria, CA) and FlowJo.

25For quantifying cell numbers expressing a marker or an area of positive cells, equal numbers of optical stacks were taken per specimen, collapsed, and quantified using Automeasure in AxioVision (Zeiss, Jena, Germany) or manually. For quantification of fluorescence intensity, 7 optical stacks were acquired from the ventral surface of animals, collapsed, and values determined using the Automeasure module (Densitometric sum) in AxioVision (Zeiss). Images were acquired using an AxioImager with Apotome (Zeiss) or an LSM 700 (Zeiss).